kubernetes提供了一種親和性調(diào)度（Affinity）。它在NodeSelector的基礎(chǔ)之上的進(jìn)行了擴(kuò)展，可以通過(guò)配置的形式，實(shí)現(xiàn)優(yōu)先選擇滿(mǎn)足條件的Node進(jìn)行調(diào)度，如果沒(méi)有，也可以調(diào)度到不滿(mǎn)足條件的節(jié)點(diǎn)上，使調(diào)度更加靈活。

Affinity主要分為三類(lèi)：

• nodeAffinity(node親和性）: 以node為目標(biāo)，解決pod可以調(diào)度到哪些node的問(wèn)題
• podAffinity(pod親和性) : 以pod為目標(biāo)，解決pod可以和哪些已存在的pod部署在同一個(gè)拓?fù)溆蛑械膯?wèn)題
• podAntiAffinity(pod反親和性) : 以pod為目標(biāo)，解決pod不能和哪些已存在pod部署在同一個(gè)拓?fù)溆蛑械膯?wèn)題
  首先我們先看下nodeAffinity，同樣我們舉個(gè)例子，方便理解

nodeAffinity

首先創(chuàng)建pod.yaml
這里使用的是 requiredDuringSchedulingIgnoredDuringExecution 硬相關(guān)，沒(méi)有匹配到，則返回失敗

apiVersion: v1
kind: Pod
metadata:
  name: nodeaffinity #pod名稱(chēng)
  namespace: dev
spec:
  containers:
  - name: nginx
    image: nginx: latest
  affinity:  #親和性設(shè)置
    nodeAffinity: #設(shè)置node親和性
      requiredDuringSchedulingIgnoredDuringExecution: # 硬限制
        nodeSelectorTerms:
        - matchExpressions: # 匹配env的值在["aa","bb"]中的標(biāo)簽
          - key: nodeenv
            operator: In
            values: ["aa","bb"]

生成yaml對(duì)應(yīng)的pod

# 創(chuàng)建pod
[root@k8s-master01 ~]# kubectl create -f pod.yaml
pod/nodeaffinity created

# 查看pod狀態(tài) （運(yùn)行失?。?[root@k8s-master01 ~]# kubectl get pods nodeaffinity -n dev -o wide
NAME                        READY   STATUS    RESTARTS   AGE   IP       NODE    ...... 
nodeaffinity   0/1     Pending   0          6s   <none>   <none>  ......

# 查看Pod的詳情
# 發(fā)現(xiàn)調(diào)度失敗，提示node選擇失敗
[root@k8s-master01 ~]# kubectl describe pod nodeaffinity -n dev
......


#接下來(lái)，停止pod
[root@k8s-master01 ~]# kubectl delete -f pod.yaml
pod "nodeaffinity" deleted

# 修改文件，將values: ["aa","bb"]------> ["pro","pod"]
[root@k8s-master01 ~]# vim pod.yaml

# 再次啟動(dòng)
[root@k8s-master01 ~]# kubectl create -f pod.yaml
pod/nodeaffinity created

# 此時(shí)查看，發(fā)現(xiàn)調(diào)度成功，已經(jīng)將pod調(diào)度到了node1上
[root@k8s-master01 ~]# kubectl get pods nodeaffinity -n dev -o wide
NAME                        READY   STATUS    RESTARTS   AGE   IP            NODE  ...... 
nodeaffinity  1/1     Running   0          11s   10.200.1.9   node1 ......

接下來(lái)再演示一下requiredDuringSchedulingIgnoredDuringExecution
同樣先創(chuàng)建pod.yaml

apiVersion: v1
kind: Pod
metadata:
  name: nodeaffinity
  namespace: dev
spec:
  containers:
  - name: nginx
    image: nginx:latest
  affinity:  #親和性設(shè)置
    nodeAffinity: #設(shè)置node親和性
      preferredDuringSchedulingIgnoredDuringExecution: # 軟限制 優(yōu)先匹配設(shè)定的規(guī)則，沒(méi)有匹配也可以運(yùn)行
      - weight: 1
        preference:
          matchExpressions: # 匹配env的值在["dev","test"]中的標(biāo)簽(當(dāng)前環(huán)境沒(méi)有)
          - key: nodeenv
            operator: In
            values: ["dev","test"]

# 創(chuàng)建pod
[root@k8s-master01 ~]# kubectl create -f pod.yaml
pod/nodeaffinity created

# 查看pod狀態(tài) （運(yùn)行成功），即使沒(méi)有這樣標(biāo)簽的node
[root@k8s-master01 ~]# kubectl get pod nodeaffinity -n dev
NAME                         READY   STATUS    RESTARTS   AGE
nodeaffinity  1/1     Running   0          40s

NodeAffinity規(guī)則設(shè)置的注意事項(xiàng)：
    1 如果同時(shí)定義了nodeSelector和nodeAffinity，那么必須兩個(gè)條件都得到滿(mǎn)足，Pod才能運(yùn)行在指定的Node上
    2 如果nodeAffinity指定了多個(gè)nodeSelectorTerms，那么只需要其中一個(gè)能夠匹配成功即可
    3 如果一個(gè)nodeSelectorTerms中有多個(gè)matchExpressions ，則一個(gè)節(jié)點(diǎn)必須滿(mǎn)足所有的才能匹配成功
    4 如果一個(gè)pod所在的Node在Pod運(yùn)行期間其標(biāo)簽發(fā)生了改變，不再符合該P(yáng)od的節(jié)點(diǎn)親和性需求，則系統(tǒng)將忽略此變化

PodAffinity

PodAffinity主要實(shí)現(xiàn)以運(yùn)行的Pod為參照，實(shí)現(xiàn)讓新創(chuàng)建的Pod跟參照pod在一個(gè)區(qū)域的功能,前面是以Node為參照。
同樣首先創(chuàng)建一個(gè)參照Pod yaml

apiVersion: v1
kind: Pod
metadata:
  name: pod-target
  namespace: dev
  labels:
    podenv: pro #設(shè)置標(biāo)簽
spec:
  containers:
  - name: nginx
    image: nginx: latest
  nodeName: node1 # 將目標(biāo)pod名確指定到node1上

# 啟動(dòng)目標(biāo)pod
[root@k8s-master01 ~]# kubectl create -f pod-target.yaml
pod/pod-target created

# 查看pod狀況
[root@k8s-master01 ~]# kubectl get pods  pod-target -n dev
NAME                     READY   STATUS    RESTARTS   AGE
pod-target   1/1     Running   0          4s

2）創(chuàng)建pod-podaffinity-required.yaml，內(nèi)容如下：

apiVersion: v1
kind: Pod
metadata:
  name: pod-podaffinity-required
  namespace: dev
spec:
  containers:
  - name: nginx
    image: nginx: latest
  affinity:  #親和性設(shè)置
    podAffinity: #設(shè)置pod親和性
      requiredDuringSchedulingIgnoredDuringExecution: # 硬限制 同樣還有軟連接效果類(lèi)似，只是即使未匹配依然可以運(yùn)行
      - labelSelector:
          matchExpressions: # 匹配env的值在["aa","bb"]中的標(biāo)簽
          - key: podenv
            operator: In
            values: ["aa","bb"]
        topologyKey: kubernetes.io/hostname

# 啟動(dòng)pod
[root@k8s-master01 ~]# kubectl create -f pod-podaffinity-required.yaml
pod/pod-podaffinity-required created

# 查看pod狀態(tài)，發(fā)現(xiàn)未運(yùn)行
[root@k8s-master01 ~]# kubectl get pods pod-podaffinity-required -n dev
NAME                       READY   STATUS    RESTARTS   AGE
pod-podaffinity-required   0/1     Pending   0          9s

# 查看詳細(xì)信息
[root@k8s-master01 ~]# kubectl describe pods pod-podaffinity-required  -n dev
......
Events:
  Type     Reason            Age        From               Message
  ----     ------            ----       ----               -------
  Warning  FailedScheduling  <unknown>  default-scheduler  0/3 nodes are available: 2 node(s) didn't match pod affinity rules, 1 node(s) had taints that the pod didn't tolerate.

# 接下來(lái)修改  values: ["aa","bb"]----->values:["pro","pod"]
[root@k8s-master01 ~]# vim pod-podaffinity-required.yaml

# 然后重新創(chuàng)建pod，查看效果
[root@k8s-master01 ~]# kubectl delete -f  pod-podaffinity-required.yaml
pod "pod-podaffinity-required" deleted
[root@k8s-master01 ~]# kubectl create -f pod-podaffinity-required.yaml
pod/pod-podaffinity-required created

# 發(fā)現(xiàn)此時(shí)Pod運(yùn)行正常
[root@k8s-master01 ~]# kubectl get pods pod-podaffinity-required -n dev
NAME                       READY   STATUS    RESTARTS   AGE   LABELS
pod-podaffinity-required   1/1     Running   0          6s    <none>

podAntiAffinity

這個(gè)與podAAffinity 使用方式完全一致，只是效果相反，不再贅述。